It looks very relevant for our discussion. Had to get some LLM help translating it into TOC language.
Here is the article translated into Theory of Constraints language.
The article’s basic point
In TOC, you ask:
What is the goal?
What is the system?
What is the constraint?
What should we change?
What should we change to?
How do we cause the change?
This article is about what happens when a system is trying to pursue a big societal mission — climate transition, circular economy, public health, social transformation, etc. — but the actors do not agree what kind of problem they are dealing with.
The article calls this a directionality challenge: the difficulty of giving a system a useful direction when the problem is complex, uncertain, contested, and interpreted differently by different stakeholders. It defines “directionality heuristics” as context-sensitive rules of thumb that guide how actors interpret a problem and decide what kind of response is appropriate.
In TOC terms:
The article is about the danger of applying the wrong improvement logic to the wrong kind of constraint.
The TOC translation
TOC often assumes that, with enough thinking, we can identify the system constraint and focus improvement there.
This article says: yes, but in wicked societal systems there is an upstream issue:
People may not even agree what kind of system-state they are in, what kind of constraint they face, or what kind of action logic is appropriate.
So before asking, “Where is the constraint?”, the article asks:
What kind of problem-space do actors believe they are in?
That belief shapes what they think the constraint is.
The five problem domains
The article borrows from the Cynefin framework. You do not need to know Cynefin; just think of it as a way of sorting problems by how knowable the cause-effect relationships are.
1. Clear problems
Cause and effect are obvious.
TOC analogy: a simple process constraint. You know what is wrong. The machine is down, the form is missing, the rule is unclear, the data is incomplete.
The right heuristic is:
sense → categorize → respond
Example response: gather the missing data, categorize it, apply a known rule.
Throughput implication: measure completion, reliability, speed, error rates.
2. Complicated problems
Cause and effect are not obvious, but experts can analyze them.
TOC analogy: a technical constraint or policy constraint that requires analysis. You can build a Current Reality Tree, consult experts, compare alternatives, and design an injection.
The right heuristic is:
sense → analyze → respond
Example response: bring experts together, evaluate options, choose better interventions.
Throughput implication: measure quality of diagnosis, validity of analysis, effectiveness of selected interventions.
3. Complex problems
Cause and effect cannot be known reliably in advance. You only know what works after trying things.
TOC analogy: you cannot fully build the FRT from the armchair. You need multiple small injections, observation, learning, and revision.
The right heuristic is:
probe → sense → respond
Example response: run several experiments, watch what happens, reinforce what works, stop what fails.
Throughput implication: measure learning rate, experiment quality, adaptive adoption, durability of emergent patterns.
4. Chaotic problems
The system is unstable and urgent. There may not be time for analysis or participatory exploration.
TOC analogy: the plant is on fire. Do not hold a strategy workshop. Stabilize the system first.
The right heuristic is:
act → sense → respond
Example response: remove a destabilizing rule, stop harm, create minimal order, then learn.
Throughput implication: measure stabilization, harm reduction, restored capacity to think and act.
5. Nescience
This is the article’s most interesting category.
Nescience means: actors do not know which domain they are in. They do not know whether the problem is clear, complicated, complex, or chaotic. Therefore, they do not know which action logic to use. The article distinguishes confusion, conscious isomorphism, and aporia: confusion means not knowing that one does not know; conscious isomorphism means knowing one does not know but not tolerating it, so one imitates familiar approaches; aporia means knowing and tolerating not-knowing, which enables more reflexive and adaptive action.
TOC analogy:
We do not yet know whether we are dealing with a physical constraint, a policy constraint, a market constraint, a measurement problem, a paradigm conflict, or an unstable crisis.
So the immediate constraint may be not the operational bottleneck, but the system’s inability to make sense of the bottleneck.
Why this matters
The article’s Proposition 2 says that different degrees of wickedness call for different directionality heuristics. Trying to tame wickedness with planning-based approaches can be as damaging as using “wickedness” as an excuse for inaction. Misaligned heuristics can undermine effectiveness and legitimacy.
In TOC language:
A solution that is appropriate for one type of constraint can become an undesirable effect when applied to another type of constraint.
For example:
- If the problem is clear, endless dialogue wastes time.
- If the problem is complicated, shallow trial-and-error may waste resources.
- If the problem is complex, expert planning can create false certainty.
- If the problem is chaotic, experiments may be too slow.
- If the problem is nescient, premature metrics can lock the system into the wrong model.
The article’s warning about measurement
This is where the article connects most strongly to throughput.
In TOC, throughput measurement is powerful because it directs attention toward the system goal.
But the article warns that in complex or nescient systems, measurement can become a false steering mechanism. Actors may use familiar monitoring and evaluation methods from old policy frames, such as model predictions, cost-benefit analyses, risk assessments, or easily quantifiable targets, even when those methods do not fit the actual problem. The article says this can bias policy toward what can be quantified and exclude important aspects that cannot yet be measured well.
TOC translation:
A throughput metric is not just a gauge. It is also a steering wheel.
If the metric fits the system, it helps exploit the constraint.
If the metric does not fit the system, it becomes part of the constraint.
How this applies to throughput
For a familiar TOC setting, throughput might be relatively clear: sales minus truly variable costs, units shipped, goal-units completed, patients treated, cases resolved, etc.
But for a transformation movement, throughput is harder. The previous 2R tree defined throughput as durable adoption of wiser views, values, practices, and social forms. That is a good start, and the tree already distinguishes throughput from activity, reach, content, attendance, money, or busyness.
This article adds a crucial refinement:
Throughput must be defined according to the kind of problem-domain the work is in.
So instead of asking only:
What is our throughput?
you ask:
What kind of problem are we working on, and what would throughput mean for that kind of problem?
A TOC-friendly table
| Problem type |
TOC-like situation |
Bad throughput metric |
Better throughput question |
| Clear |
Known process problem |
Number of meetings about it |
Are known tasks completed reliably? |
| Complicated |
Expert diagnosis needed |
Number of opinions gathered |
Has expert analysis improved the intervention? |
| Complex |
Emergent transformation |
Fixed target hit rate |
Are probes producing learning and durable adoption? |
| Chaotic |
Urgent instability |
Quality of long-term plan |
Is harm reduced and stability restored? |
| Nescient |
Unknown problem type |
Premature KPI |
Are we becoming better able to sense what kind of problem this is? |
What the article adds to TOC
The article does not reject TOC. It adds a prior diagnostic layer.
Classic TOC question:
What is the constraint?
Article’s prior question:
What kind of situation are we in, and what kind of heuristic is appropriate for identifying and acting on the constraint?
In a clear or complicated system, the constraint can often be identified through observation and analysis.
In a complex system, the constraint may only become visible through probes.
In a chaotic system, the first “constraint” may be instability itself.
In nescience, the constraint may be the group’s lack of awareness or tolerance of not-knowing.
The article in one TOC sentence
The article argues that in mission-driven, wicked-problem systems, the first constraint may be misaligned sense-making: actors apply clear, complicated, complex, or chaotic problem-solving heuristics without first understanding which kind of problem they are facing.
Why it matters for the 2R trees
The current 2R trees say the group should define throughput, build a conversion pathway, create practice architecture, form pockets, document patterns, and review constraints monthly.
The article says: good, but before defining throughput too rigidly, ask:
Is “defining throughput” itself a clear, complicated, complex, chaotic, or nescient problem?
For 2R, it is probably nescient moving into complex.
That means the first version of throughput should be provisional, mixed-method, and revisable. The existing FRT already warns that throughput metrics can become reductive and that the group needs mixed-method measurement to avoid becoming managerial and losing soul.
The article strongly supports that warning.
Practical takeaway
For someone trained in TOC, the article’s lesson is:
Do not assume every constraint can be handled with the same improvement logic.
First classify the problem-space:
- If it is clear, standardize.
- If it is complicated, analyze.
- If it is complex, experiment.
- If it is chaotic, stabilize.
- If it is nescient, cultivate awareness and tolerance of not-knowing before locking in direction.
Then define throughput accordingly.
So for 2R, a better throughput definition might be:
Throughput is not only the rate of durable adoption. It is also the rate at which the movement increases its capacity to orient wisely under complexity, uncertainty, contestation, and not-knowing.